Online torque estimation of internal combustion engines using neural networks

An internal combustion engine torque estimation method based on artificial neural networks has been proposed in this paper. Besides, an investigation has been done to show neural networks ability to detect faults like misfire, O₂ sensor trouble, MAP sensor trouble and fuel injector trouble. Unlike most other methods there is no need to auxiliary add-on systems to collect data and all of the estimation procedure takes place via available data on the on-board diagnostic data line. In other similar methods some additional sensors should be used. This method has been tested on a real experimental XU7; Peugeot engine connected to a dynamometer. Estimated torque values and determined faults have been compared with the real experimental data.